specialist-engineering-backend-architect
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interpolates user-supplied text directly into the system prompt of a spawned sub-agent without protective boundary markers or sanitization, which could allow user input to influence the agent's internal instructions.
- Ingestion points: The
{the user's question}placeholder in theAgenttool call inSKILL.mdis populated with raw user data. - Boundary markers: No delimiters, such as XML tags or dedicated section headers, are used to isolate user input from the sub-agent's instructions.
- Capability inventory: The sub-agent is designed to generate complex technical artifacts, including database schemas, system architecture specifications, and API implementations.
- Sanitization: The skill does not perform any escaping or validation on the input string before it is interpolated into the prompt.
Audit Metadata